constructed using a Cox elastic net algorithm with repeated crossvalidation (CV).Across 20 iterations, ACs were ranked by selection frequency and mean absolute coefficient values.After removing highly correlated features, prediction models were developed using the clinical covariates above combined with the top n metabolites.Model performance was assessed using 100 repeated 5-fold CVs, and the combination of metabolites with the highest concordance index (Cindex) was selected.Results: Among the 97 participants, 60.3% had FSGS, 64.0% were adults, and 63.2% were male.Over a median follow-up of 51.0 months, 49.5% achieved CR.Metabolomics data included 21 plasma and 25 urine ACs with diverse chain lengths, degrees of saturation, and subclasses (Figure 1).Several ACs consistently showed significant associations (p < 0.05) with CR in both plasma and urine.Models incorporating ACs markedly improved prediction of CR compared with models using only clinical covariates.The optimal models identified eight ACs with the strongest predictive value for CR in plasma-CAR(DC(10:0)), CAR( 22:4(OH)), CAR(DC9:0), CAR(DC9:1), CAR(12:1(OH)), CAR(17:2), and CAR(14:2(3,5))-and eight in urine-CAR(10:2), CAR(21:4(OH)), CAR(DC5:1), CAR(7:0(OH)), CAR(19:2), CAR(DC11:2), CAR(14:2(3,5)), and CAR(17:2) (Figure 2).Conclusion: Models incorporating ACs significantly improved prediction of proteinuria remission in FSGS/MCD, suggesting that ACs represent promising biomarkers for treatment response.Further mechanistic studies on AC metabolism in the kidney may provide deeper insights into disease mechanisms and potential therapeutic targets for FSGS/MCD.I have no potential conflict of interest to disclose.I did not use generative AI and AI-assisted technologies in the writing process.
Lim et al. (Wed,) studied this question.